China
Africa
Explore chapters and articles related to this topic
D
Published in Philip A. Laplante, Comprehensive Dictionary of Electrical Engineering, 2018
differentiator differentiator a filter that performs a differentiation of the signal. Since convolution and differentiation are both linear operations, they can be performed in either order. ( f g) (x) = f (x) g(x) = f (x) g (x). Thus, instead of filtering a signal and then differentiating the result, differentiating the filter and applying it to the signal has the same effect. This filter is called a differentiator. A low-pass filter is commonly differentiated and used as a differentiator. diffracted beam diffraction that takes place when the wavelength of an incident beam is short compared to the interaction distance. Particles exhibit wave like characteristics in their passage through matter. In striking a target the incident beam scatters off nucleons. The scattered waves then combine according to the superposition principle and the peak of this scattered wave is called the diffracted beam. diffraction angle angle corresponding approximately to the rate of spreading of an electromagnetic wave that has been transmitted through an aperture; with Gaussian beams the far field half angle for a radius equal to the spot size. diffraction coefficient in the Geometric Theory of Diffraction, the coefficient that is proportional to the contribution to the scattered field due to the fringe currents near an edge or corner of a scattering target. diffraction efficiency of Bragg cell ratio of the intensity of the principal diffracted beam to the intensity of the undiffracted beam. diffraction grating an array of reflecting or transmitting lines that mutually enhance the effects of diffraction. diffraction loss loss from an electromagnetic beam due to finite aperture effects. diffraction tomography generalization of computerized tomography incorporating scattering effects. diffuse density signal that has uniform energy density, meaning that the energy flux is equal in all parts of a given region. diffuse intensity the energy scattered in all directions out of the forward or specular directions. Sometimes also called incoherent component of the intensity. diffuse multipath the result of multipath propagation observed as overlapping signal components, due to delay differences of multipath components being less than the delay resolution of the signal. Observable in the delay power spectrum as a continuous distribution of power over delay. See also multipath propagation, delay power spectrum. diffuse scattering the component of the scattering from a rough surface that is not in the specular direction. It is caused by reflections from local surfaces oriented in planes different from that of the mean surface. See also specular scattering. diffuse transmittance a transmitted signal that has uniform energy density. diffusion a region of a semiconductor into which a very high concentration of impurity has been diffused in order to substantially increase the majority carrier concentration in that region. diffusion pump second stage of the vacuum system. Hot oil showers the particles in a vacuum and creates a better vacuum. After a mechanical (roughing) pump is used to remove about 99.99% of the air in the beam tube, the remaining air can then be removed by a diffusion pump, down to about 1E-9 torr.
Low-dose CT image denoising using sparse 3d transformation with probabilistic non-local means for clinical applications
Published in The Imaging Science Journal, 2023
Dawa Chyophel Lepcha, Bhawna Goyal, Ayush Dogra
The deep learning technique has made substantial progress in the domain of LDCT image denoising in recent years [13–16]. The Residual encoder decoder convolutional neural network (RED-CNN) [17], that are inspired from deep learning, combines autoencoder, deconvolutional network and shortcut networks for LDCT denoising. In both simulated and clinical scenarios, RED-CNN obtains finer results in comparison to other approaches after patch-based training. This method has received high performance in case of noise reduction, lesion diagnosis and structural preservation. Another approach for LDCT denoising is based on quadratic autoencoder (QAE) [18], which employs quadratic neurons to build encoder decoder design. In case of image denoising as well as model efficiency, the experiment performance on Mayo LDCT dataset represents the utility and potential of quadratic autoencoder. In our apprehension, this is the first deep learning approach that has been used with novel form of neuron, and it shows great promise in the domain of medical imaging. In case of LDCT images, the two-stage residual convolutional neural network (TS-RCNN) [19] was introduced. In terms of this network, there are two key components. The stationary wavelet transform (SWT) and the perceptual loss are used in the first step of RCNN for textural denoising. The SWT is implemented on each NCDT images and then obtained four wavelet images that are considered as labels. Based on initial network results, the second step RCNN is constructed for structure enhancement through average NDCT model. Finally using inverse SWT, the denoised CT image is obtained. The CNN-based denoising which does not require any separate training data beyond the noisy reconstruction that are already available is proposed. The Noise2Inverse method [20] is one of them and it is designed specifically for computed tomography and relative inverse applications. Until now, Noise2Inverse applications are considered for 2D spatial information. In addition, an expansion for a Noise2Inverse application is carried out in domains such as space, time, and spectrum. Static and dynamic micro-tomography and X-ray diffraction tomography will benefit from this advancement. The Noise2inverse is capable for appropriate denoising and favours for significant suppression in acquisition time while retaining quality of images according to the results from real-world datasets.